stanj
Offline
• • • • • •
Upload & Sell: On
|
Nill Toulme wrote:
But it's a compromise in ways beyond just size... that's what I would like to see articulated. IOW, if a 10MP Mark IV mRAW file (I don't know if that's the actual size, just made that up) were just like a full RAW instead of it's 10MP instead of 18MP, OK, it would just be less resolution and that might be fine. But I'm hearing that's is less than that... that the compromises are more that. So... how so, and how does this weigh as against shooting jpg when we could otherwise do nicely with less resolution and a smaller file size but still want all the general benefits of RAW?...Show more →
For starters, anytime you remove data, you're removing information, quality, regardless how good it is. BR video compression, JPEG, all that. Specifically about sRaw, on the sensor you start with each pixel having one value: R, G, B. Four hardware pixels make out one full RGB pixel; the way cameras are marketed, a 16MP camera has 16 million photo sites, but actually only 4M discreet RGB values; the other ones are interpolated. This step is called debeyer, as it removes the Beyer pattern. This step is where smoothness of lines or moire is determined - and where the best noise reduction and sharpening can happen, because you have the raw pixels in front of you. This is how ACR6 is better than ACR5 - the debeyer is completely new, better, more sophisticated.
Now, sRaw gives you a debeyered image. A 16MP camera will give you a 4MP sRaw image (mRaw is an interesting tweener, but the differences in approach are mostly irrelevant). In the Canon sRaw files, each pixel has an RGB value, not just either R or G or B, like in real raw. Therefore, one may think that an sRaw image would be 1/4 the file size of a raw, yet it is about 3/4 the file size of a true raw - because it carries 3x as much information "per pixel". But we have already lost information: we carry only 3/4 of the information.
The other problem is _how_ this information was lost, or rather, how the debeyer step was performed. Take a modern day Nehalem class machine and open an image in ACR6 or your choice of raw converter. You'll wait about 2s for that. The machine draws about 60-200W power, depending on what you have in it. In contrast, your camera does say eight such conversions per second. Granted, it has special hardware for this task only, but still - shortcuts need to be taken, hardcoded decisions must be made. Your computer has more time to deal with this - it can make much more sophisticated decisions. When you have an sRaw file, this decision making is gone; all it can do is "classic" noise reduction / sharpening, on top of the debeyer. Since the step that has the most impact on image quality is performed in camera, with a technology put in stone at the time of the release of the camera, additionally using a technology that takes shortcuts, you're stuck with something that's mediocre. Maybe sufficient for your needs, but far from optimal.
It's no wonder that sRaw files are less noisy than their raw counterparts - they are 1/4 the size. But do this experiment: take a high contrast, typical night scene shot at ISO 3200, shoot it sRaw, and then raw opened in ACR6, and then shrunk down to 1/4 its size using Bicubic or Bicubic Sharper (just to keep it simple). Then decide which one is better. Unless you can create a synthetic case that trips a bug in the raw converter, the full raw shrunk will always win.
Given that sRaw files aren't proportionally smaller than raw files (they aren't 1/4 the file size), the only real benefit of sRaw files is for reasons of memory pressure during processing (the image will be 1/4 the size, which is a huge CPU saving), as well as processing time, since the debeyer step - which accounts for the lion's share of the time of a decode - is skipped.
I'm not passing judgement on people who use it - I know that Canon added it for a reason, responding to a market request. Doesn't mean the market actually understands what it's asking for, or rather getting.
|
